Model training method and device, computer equipment and readable storage medium
By displaying and customizing processing unit labels, the model training link is dynamically constructed, which solves the problem of fixed model training processing flow in the existing technology and is difficult to adapt to variable tasks, and realizes flexible and efficient model training.
Patent Information
- Application Number
- CN202510254901.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-05
- Publication Date
- 2025-05-30
AI Technical Summary
There are existing methods for model training in a workflow method. The processing flow is relatively fixed, the flexibility is poor, and it is difficult to adapt to changeable model training tasks. The workflow needs to be reconstructed when switching model training tasks.
Provide a model training method. By displaying and customizing processing unit tags, users can flexibly connect processing units and set parameter items, dynamically build model training links, and switch model training tasks by adjusting label and parameter item information.
The flexibility and efficiency of model training are realized. Users can adjust the connection order and parameter information of processing unit labels according to their needs to adapt to changeable model training tasks without rebuilding the model training link, which improves the efficiency of model training.
Smart Images
Figure CN120066349A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of artificial intelligence technology, and in particular, to a model training method, apparatus, computer device, computer-readable storage medium, and computer program product. Background Art
[0002] With the rapid increase in the amount of data, the complexity and diversity of the data make model training more complex. In order to improve the efficiency of model training, model training is carried out in the form of a workflow.
[0003] However, in the current method of model training in the form of a workflow, the processing flow of model training is relatively fixed, with poor flexibility. Moreover, when performing different model training tasks, it is necessary to reconstruct the workflow, making it difficult to adapt to the changing model training tasks. Summary of the Invention
[0004] Based on this, in view of the above technical problems, it is necessary to provide a model training method, apparatus, computer device, computer-readable storage medium, and computer program product.
[0005] In a first aspect, this application provides a model training method, including:
[0006] Display the processing unit labels corresponding to each processing unit involved in model training, and display the input connection points and output connection points included in each processing unit label, and display the parameter item setting controls included in the customizable parameter processing unit label; the customizable parameter processing unit label is a processing unit label whose parameter items can be customized by the user;
[0007] Based on the operation of the user connecting the output connection point of at least one first processing unit label to the input connection point of another first processing unit label, and / or the parameter item information set by the user in the parameter item setting control, obtain a first model training link; the first processing unit label is the processing unit label required by the user to train a first model;
[0008] Perform model training based on the first model training link to obtain a first model.
[0009] In one embodiment, the parameter item setting control includes at least one of a network structure switching control for the model to be trained, a format parameter setting control, and a function opening and closing setting control.
[0010] In one embodiment, after obtaining the first model training link, the method further includes:
[0011] Adjust the first model training link to obtain a second model training link;
[0012] Perform model training based on the second model training link to obtain a second model.
[0013] In one embodiment, adjusting the first model training link to obtain a second model training link includes:
[0014] In response to the user's operation of adding a tag, deleting a tag, or adjusting the order between tags, adjust the first model training link to obtain a second model training link;
[0015] And / or, in response to the user's operation of adjusting parameter item information, adjust the first model training link to obtain a second model training link.
[0016] In one embodiment, before displaying the processing unit tags corresponding to each processing unit involved in model training, displaying the input connection points and output connection points included in each processing unit tag, and displaying the parameter item setting controls included in the customizable parameter processing unit tag, the method further includes:
[0017] Determine the data types of each input connection point and each output connection point;
[0018] Obtain the target format standards corresponding to each data type;
[0019] According to the target format standards corresponding to each data type and the data types of the input connection points and output connection points of each processing unit tag, perform format standardization processing on the input connection points and output connection points of each processing unit tag.
[0020] In one embodiment, obtaining a first model training link based on the user's operation of connecting the output connection point of at least one first processing unit tag to the input connection point of another first processing unit tag includes:
[0021] Based on the user's operation of arranging each of the first processing unit tags according to the sequence of each processing process required for training the first model, obtain the arranged first processing unit tags;
[0022] Based on the user's operation of connecting the arranged target processing unit tags according to the input connection points and output connection points of the first processing unit tags, obtain a first model training link.
[0023] In one embodiment, the performing model training based on the first model training link to obtain a first model includes:
[0024] Based on the connection order of each first processing unit label in the first model training link, call the pre-stored encapsulated code segments corresponding to each first processing unit label to obtain the model training code corresponding to the first model training link;
[0025] Perform model training based on the model training code corresponding to the first model training link to obtain a first model.
[0026] In a second aspect, the present application also provides a model training device, including:
[0027] A processing unit label display module, configured to display the processing unit labels corresponding to each processing unit involved in model training, and display the input connection points and output connection points included in each processing unit label, and display the parameter item setting controls included in the customizable parameter processing unit label; the customizable parameter processing unit label is a processing unit label whose parameter items can be customized by the user;
[0028] A first model training link acquisition module, configured to obtain a first model training link based on the operation of the user connecting the output connection point of at least one first processing unit label to the input connection point of another first processing unit label, and / or, the parameter item information set by the user in the parameter item setting control; the first processing unit label is the processing unit label required for the user to train a first model;
[0029] A first model training module, configured to perform model training based on the first model training link to obtain a first model.
[0030] In a third aspect, the present application also provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor executes the above method.
[0031] In a fourth aspect, the present application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and the computer program is executed by a processor to perform the above method.
[0032] In a fifth aspect, the present application also provides a computer program product. The computer program product includes a computer program, and the computer program is executed by a processor to perform the above method.
[0033] The above model training method, device, computer device, computer-readable storage medium, and computer program product display the processing unit labels corresponding to each processing unit involved in model training, display the input connection points and output connection points included in each processing unit label, and display the parameter item setting controls included in the customizable parameter processing unit label; the customizable parameter processing unit label is the processing unit label for which the user can customize parameter items; based on the operation of the user connecting the output connection point of at least one first processing unit label to the input connection point of another first processing unit label, and / or the parameter item information set by the user in the parameter item setting control, a first model training link is obtained; the first processing unit label is the processing unit label required for the user to train the first model; model training is performed based on the first model training link to obtain the first model. In this application, based on the operation of the user connecting the output connection point of at least one first processing unit label to the input connection point of another first processing unit label, and / or the parameter item information set by the user in the parameter item setting control, a first model training link is obtained to obtain the first model, and model training is carried out in a visual manner, with convenient operation and strong readability. The connection order of the processing unit labels and the parameter item information can be changed according to the requirements of model training, thereby changing the processing flow of model training, and the flexibility is relatively high. Moreover, when switching model training tasks, there is no need to reconstruct the model training link, and the switching of model training tasks can be achieved through operations such as label addition, label deletion, order adjustment between labels, or parameter item information adjustment, which can adapt to changing model training tasks and improve the efficiency of model training. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments of the present application or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other related drawings can be obtained without creative efforts based on these drawings.
[0035] Figure 1 It is an application environment diagram of the model training method in an embodiment;
[0036] Figure 2 It is a flowchart of the model training method in an embodiment;
[0037] Figure 3 It is a schematic diagram of the network structure switching control in an embodiment;
[0038] Figure 4 It is a schematic diagram of the format parameter setting control in an embodiment;
[0039] Figure 5Schematic diagram of a function opening and closing setting control in an embodiment;
[0040] Figure 6 Schematic diagram of a partial link of a first model training link in an embodiment;
[0041] Figure 7 Schematic diagram of a partial link of a second model training link in an embodiment;
[0042] Figure 8 Structural block diagram of a model training device in an embodiment;
[0043] Figure 9 Internal structure diagram of a computer device in an embodiment. Detailed implementation manners
[0044] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0045] The embodiment of the present application provides a model training method, and this embodiment can be executed by a computer device. As Figure 1 shown, the computer device can display the processing unit labels corresponding to each processing unit involved in model training; the user can, based on the processing unit labels displayed by the computer device, connect the output connection point of at least one first processing unit label to the input connection point of another first processing unit label, and / or set parameter item information in the parameter item setting control in the custom parameter processing unit label to obtain a first model training link. The computer device can perform model training based on this first model training link to obtain a first model. It can be understood that the computer device can be implemented through a server or through a terminal. In this embodiment, the method includes Figure 2 the steps shown:
[0046] Step S201: Display the processing unit labels corresponding to each processing unit involved in model training, and display the input connection points and output connection points included in each processing unit label, and display the parameter item setting control included in the custom parameter processing unit label; the custom parameter processing unit label is a processing unit label whose parameter items can be customized by the user.
[0047] The computer device can generate the processing unit labels corresponding to each processing unit according to each processing unit involved in model training. After the processing unit labels corresponding to each processing unit involved in model training are prepared, subsequent model training will no longer require adjustment at the code layer. When adding or deleting a certain processing process of model training, only the corresponding processing unit label needs to be added or deleted.
[0048] The computer device can display the processing unit labels corresponding to each processing unit involved in model training to the user, display the input connection points and output connection points included in each processing unit label, and display the parameter item setting controls included in the customizable parameter processing unit label, so that the user can obtain the first model training link.
[0049] Step S202: Based on the operation of the user connecting the output connection point of at least one first processing unit label to the input connection point of another first processing unit label, and / or the parameter item information set by the user in the parameter item setting control, obtain the first model training link; the first processing unit label is the processing unit label required for the user to train the first model.
[0050] The user can determine the first processing unit label based on the processing unit labels required for training the first model.
[0051] When the processing unit labels required for training the first model do not include the customizable parameter processing unit label, the user can, based on the processing unit labels displayed by the computer device, the input connection points and output connection points included in each processing unit label, the parameter item setting controls included in the customizable parameter processing unit label, and each processing process required for training the first model, connect the output connection point of at least one first processing unit label to the input connection point of another first processing unit label, so that the computer device can obtain the first model training link. Among them, the model training link can also be called the model training workflow, and the processing unit label can be called the basic node.
[0052] When the processing unit labels required for training the first model include the customizable parameter processing unit label, the user can, based on the processing unit labels displayed by the computer device, the input connection points and output connection points included in each processing unit label, the parameter item setting controls included in the customizable parameter processing unit label, and each processing process required for training the first model, connect the output connection point of at least one first processing unit label to the input connection point of another first processing unit label, and set the parameter item information corresponding to the first model training in the parameter item setting control, so that the computer device can obtain the first model training link.
[0053] Step S203: Perform model training based on the first model training link to obtain the first model.
[0054] The computer device can perform model training based on the first model training link to obtain the first model.
[0055] In the above model training method, based on the operation of the user connecting the output connection point of at least one first processing unit label to the input connection point of another first processing unit label, and / or the parameter item information set by the user in the parameter item setting control, a first model training link is obtained to obtain a first model. The model training is carried out in a visual manner, with convenient operation and strong readability. The connection order of the processing unit labels and the parameter item information can be changed according to the needs of model training, so as to change the processing flow of model training, and the flexibility is relatively high. Moreover, when switching the model training task, there is no need to reconstruct the model training link, and the switching of the model training task can be achieved through operations such as label addition, label deletion, order adjustment between labels, or parameter item information adjustment, which can adapt to various model training tasks. For example, it is applicable to the training task of an end-to-end voice chat model, and can improve the efficiency of model training. In addition, when troubleshooting model problems, only the input and output of each processing unit label need to be viewed to quickly locate the problem.
[0056] In one embodiment, the parameter item setting control includes at least one of a network structure switching control for the model to be trained, a format parameter setting control, and a function on / off setting control.
[0057] The parameter item setting control may include at least one of a network structure switching control for the model to be trained, a format parameter setting control, and a function on / off setting control.
[0058] The network structure switching control can be as Figure 3 shown, and the switching of the network structure of the processing unit label can be achieved through a drop-down menu. Among them, the Chinese of LanguageModel is voice model, the Chinese of audio is audio, indicating that the data types of the input connection point and the output connection point of this network structure switching control are both audio types. The Chinese of token is token, the Chinese of model_siructure is model structure, the Chinese of num_layer is number of layers, the Chinese of input_dims is input dimension, and the Chinese of output_dims is output dimension.
[0059] The format parameter setting control can be as Figure 4As shown, where the Chinese of ChunkTokenStream is "fragment token stream", and the Chinese of text tokens is "text token", indicating that the data type of one of the input connection points of this format parameter setting control is the text token type; the Chinese of audio tokens is "audio token", indicating that the data type of the other input connection point of this format parameter setting control is the audio token type; the Chinese of mixed tokens is "mixed token", indicating that the data type of the output connection point of this format parameter setting control is the mixed token type; the Chinese of text chunk len is "text segment length", and 13 is the format parameter corresponding to text chunk len, which can be set by the user; the Chinese of audio chunklen is "audio segment length", and 26 is the format parameter corresponding to audio chunk len, which can be set by the user.
[0060] The function on / off setting control can be as Figure 5 shown, where the Chinese of AudioPreprocess is "audio preprocessing", and the Chinese of audio is "audio", indicating that the data types of both the input connection point and the output connection point of this function on / off setting control are audio types; the Chinese of min len is "minimum length", the Chinese of max len is "maximum length", the Chinese of denoise is "denoising", the Chinese of enhance is "enhancing", and the Chinese of enable is "enable", and the user can set whether to enable the corresponding audio preprocessing means.
[0061] In this embodiment, the parameter item setting control may include at least one of a network structure switching control for the model to be trained, a format parameter setting control, and a function on / off setting control, so that the switching of the model training task can be adjusted according to the parameter item information, and the efficiency of model training can be improved.
[0062] In one embodiment, after obtaining the first model training link, the method provided by this application further includes: adjusting the first model training link to obtain a second model training link; performing model training based on the second model training link to obtain a second model.
[0063] When the model training task is switched from the first model training task to the second model training task, there is no need to reconstruct the second model training link. The computer device can, based on the operations of the user to adjust the first model training link according to each processing process required for training the second model, obtain the second model training link; the computer device can perform model training based on the second model training link to obtain a second model.
[0064] In this embodiment, when switching the model training task, there is no need to reconstruct the model training link. By adjusting the existing model training link, the switching of the model training task can be achieved. When the model training task changes, the parts that do not need to be changed currently (such as processing unit labels, parameter item information, and data flow) are retained to the greatest extent, simplifying the workload required for adjusting the model training task, adapting to the changing model training tasks, and improving the efficiency of model training.
[0065] In one of the embodiments, the first model training link is adjusted to obtain a second model training link. The specific steps are as follows: in response to the user's operation of adding a label, deleting a label, or adjusting the order between labels, the first model training link is adjusted to obtain a second model training link; and / or, in response to the user's operation of adjusting parameter item information, the first model training link is adjusted to obtain a second model training link.
[0066] When the user determines, based on the processing processes required for training the first model and the processing processes required for training the second model, that the difference between the processing processes required for training the second model and the processing processes required for training the first model is that a label needs to be added, a label needs to be deleted, or the order between labels needs to be adjusted, the user can determine to perform an operation of adding a label, deleting a label, or adjusting the order between labels on the first model training link and execute the corresponding operation of adding a label, deleting a label, or adjusting the order between labels. The computer device can, in response to the operation of adding a label, deleting a label, or adjusting the order between labels on the first model training link performed by the user, adjust the first model training link to obtain a second model training link.
[0067] Exemplarily, a partial link of the first model training link is as Figure 6 shown, where the Chinese of TextNormalization is text regularization processing, the Chinese of TextTokenizer is text tokenizer, and the Chinese of AudioTokenizer is audio tokenizer.
[0068] Based on the processing procedures required for training the first model and the processing procedures required for training the second model, the user can determine that the difference between the processing procedures required for training the second model and the processing procedures required for training the first model is that: the processing procedure for the input audio data in the first model is changed to input text data for speech synthesis, and the generated audio data is sent to the subsequent processing procedures. Then the user can determine that the difference between the processing procedures required for training the second model and the processing procedures required for training the first model is that new tags need to be added and tags need to be deleted. At this time, the user can determine to perform an operation of adding tags, an operation of deleting tags, and an operation of adjusting the order between tags on the training link of the first model and execute the corresponding operations of adding tags, deleting tags, and adjusting the order between tags, and adjust the training link of the first model to obtain the training link of the second model, where a part of the training link of the second model is as Figure 7 shown. Among them, the Chinese of TTS is speech synthesis, the full English name is Text-to-Speech, and the Chinese of roles is role.
[0069] For the above process of adjusting the training link of the first model, no code modification is required. Only an operation of adding tags, an operation of deleting tags, and an operation of adjusting the order between tags need to be performed on the training link of the first model. The specific steps of performing an operation of adding tags, an operation of deleting tags, and an operation of adjusting the order between tags on the training link of the first model are as follows: Add a speech synthesis processing unit tag to the training link of the first model, delete the audio preprocessing unit tag in the training link of the first model, and connect the output connection point of the text regularization processing unit tag in the training link of the first model to the input connection point of the speech synthesis processing unit tag.
[0070] When the user, based on the processing procedures required for training the first model and the processing procedures required for training the second model, determines that the difference between the processing procedures required for training the second model and the processing procedures required for training the first model is that the parameter item information needs to be adjusted, the user can determine to perform a parameter item information adjustment operation on the training link of the first model and execute the corresponding parameter item information adjustment operation. The computer device can respond to the parameter item information adjustment operation performed by the user on the training link of the first model and adjust the training link of the first model to obtain the training link of the second model.
[0071] When the user determines that the differences between the processing procedures required for training the second model and those required for training the first model lie in the need to add labels, delete labels, or adjust the order of labels, and the need to adjust parameter item information, based on the processing procedures required for training the first model and those required for training the second model, the user can determine to perform a label addition operation, a label deletion operation, or an order adjustment operation between labels on the first model training link, and execute the corresponding label addition operation, label deletion operation, or order adjustment operation between labels. Then, perform a parameter item information adjustment operation on the first model training link and execute the corresponding parameter item information adjustment operation. The computer device can respond to the adjustment operation performed by the user on the first model training link and adjust the first model training link to obtain the second model training link.
[0072] In this embodiment, the computer device can respond to the user's label addition operation, label deletion operation, or order adjustment operation between labels, adjust the first model training link to obtain the second model training link; and / or, respond to the user's parameter item information adjustment operation, adjust the first model training link to obtain the second model training link. The switching of the model training task can be realized by performing a label addition operation, a label deletion operation, an order adjustment operation between labels, or a parameter item information adjustment operation on the existing model training link. The adjustment of the model training link can be carried out in a visual manner, which can adapt to changing model training tasks and improve the efficiency of model training.
[0073] In one of the embodiments, before displaying the processing unit labels corresponding to the processing units involved in model training, displaying the input connection points and output connection points included in each processing unit label, and displaying the parameter item setting controls included in the customizable parameter processing unit label, the method provided in this application further includes: determining the data types of each input connection point and each output connection point; obtaining the target format standards corresponding to each data type; and performing format standardization processing on the input connection points and output connection points of each processing unit label according to the target format standards corresponding to each data type and the data types of the input connection points and output connection points of each processing unit label.
[0074] When the computer device generates the processing unit labels corresponding to the processing units involved in model training, it needs to determine the data types of each input connection point and each output connection point of each processing unit label; perform format standardization processing on the input connection points and output connection points of each processing unit label according to the target format standards corresponding to each data type and the data types of the input connection points and output connection points of each processing unit label, so as to unify the formats of the input connection points and output connection points of each processing unit label.
[0075] In one embodiment, based on the operation of the user connecting the output connection point of at least one first processing unit label to the input connection point of another first processing unit label, a first model training link is obtained. The specific steps are as follows: Based on the operation of the user arranging the first processing unit labels according to the sequence of each processing process required for training the first model, the arranged first processing unit labels are obtained; based on the operation of the user connecting the arranged target processing unit labels according to the input connection point and output connection point of the first processing unit label, a first model training link is obtained.
[0076] The user can arrange the first processing unit labels according to the sequence of each processing process required for training the first model, and connect the arranged target processing unit labels according to the input connection point and output connection point of the first processing unit label.
[0077] Among them, the output connection point of any first processing unit label is connected to the input connection point of the first processing unit label arranged one position after it; the data types of the connected output connection point and input connection point are the same.
[0078] The computer can obtain the arranged first processing unit labels based on the operation of the user arranging the first processing unit labels according to the sequence of each processing process required for training the first model; based on the operation of the user connecting the arranged target processing unit labels according to the input connection point and output connection point of the first processing unit label, a first model training link is obtained.
[0079] In this embodiment, based on the operation of the user arranging the first processing unit labels according to the sequence of each processing process required for training the first model, the arranged first processing unit labels are obtained; based on the operation of the user connecting the arranged target processing unit labels according to the input connection point and output connection point of the first processing unit label, a first model training link is obtained, so that the context of the first model training link is concise and clear.
[0080] In one embodiment, based on the first model training link, model training is performed to obtain a first model. The specific steps are as follows: Based on the connection sequence of each first processing unit label in the first model training link, the pre-stored encapsulated code segments corresponding to each first processing unit label are called to obtain the model training code corresponding to the first model training link; based on the model training code corresponding to the first model training link, model training is performed to obtain a first model.
[0081] The computer device may call the pre-stored encapsulated code segments corresponding to each first processing unit label based on the connection order of the first processing unit labels in the first model training link to obtain the model training code corresponding to the first model training link; and may perform model training based on the model training code corresponding to the first model training link to obtain the first model.
[0082] In this embodiment, by calling the pre-stored encapsulated code segments corresponding to each first processing unit label based on the connection order of the first processing unit labels in the first model training link to obtain the model training code corresponding to the first model training link for model training to obtain the first model, the efficiency of model training can be improved.
[0083] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are displayed sequentially as indicated by the arrows, these steps do not necessarily need to be executed sequentially in the order indicated by the arrows. Unless specifically stated in this article, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages do not necessarily need to be executed at the same time, but can be executed at different times. The execution order of these steps or stages does not necessarily need to be sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.
[0084] Based on the same inventive concept, an embodiment of the present application further provides a model training device for implementing the model training method described above. The implementation solutions for solving problems provided by this device are similar to the implementation solutions described in the above method. Therefore, the specific limitations in one or more embodiments of the model training device provided below can refer to the limitations on the model training method in the above text and will not be repeated here.
[0085] In an exemplary embodiment, as Figure 8 shown, a model training device is provided, where:
[0086] The processing unit label display module 801 is configured to display the processing unit labels corresponding to each processing unit involved in model training, and display the input connection points and output connection points included in each processing unit label, and display the parameter item setting controls included in the custom parameter processing unit label; the custom parameter processing unit label is a processing unit label for which the user can customize parameter items.
[0087] The first model training link acquisition module 802 is configured to obtain a first model training link based on the operation of the user connecting the output connection point of at least one first processing unit label to the input connection point of another first processing unit label, and / or the parameter item information set by the user in the parameter item setting control; the first processing unit label is the processing unit label required by the user to train the first model.
[0088] The first model training module 803 is configured to perform model training based on the first model training link to obtain a first model.
[0089] In one embodiment, the parameter item setting control includes at least one of a network structure switching control for the model to be trained, a format parameter setting control, and a function on / off setting control.
[0090] In one embodiment, after obtaining the first model training link, the device further includes a second model acquisition module, configured to: adjust the first model training link to obtain a second model training link; perform model training based on the second model training link to obtain a second model.
[0091] In one embodiment, the second model acquisition module is further configured to: in response to the user's label addition operation, label deletion operation, or label order adjustment operation, adjust the first model training link to obtain a second model training link; and / or in response to the user's parameter item information adjustment operation, adjust the first model training link to obtain a second model training link.
[0092] In one embodiment, the device further includes a format standardization processing module, configured to: determine the data types of each input connection point and each output connection point; obtain the target format standard corresponding to each data type; perform format standardization processing on the input connection point and output connection point of each processing unit label according to the target format standard corresponding to each data type and the data types of the input connection point and output connection point of each processing unit label.
[0093] In one embodiment, the first model training link acquisition module 802 is further configured to: based on the operation of the user arranging each of the first processing unit labels according to the sequence of each processing process required for training the first model, obtain the arranged first processing unit labels; based on the operation of the user connecting the arranged target processing unit labels according to the input connection points and output connection points of the first processing unit labels, obtain the first model training link.
[0094] In one embodiment, the first model training module 803 is further configured to: based on the connection order of each first processing unit tag in the first model training link, call the pre-stored encapsulated code segments corresponding to each first processing unit tag to obtain the model training code corresponding to the first model training link; perform model training based on the model training code corresponding to the first model training link to obtain a first model.
[0095] Each module in the above model training device can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.
[0096] In an exemplary embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 9 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the data of the embodiments of the model training method. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a model training method.
[0097] Those skilled in the art can understand that Figure 9 the structure shown in
[0098] is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.
[0099] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the foregoing method embodiments are implemented.
[0100] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the foregoing method embodiments are implemented.
[0101] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.
[0102] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, artificial intelligence (AI) processors, etc., without limitation.
[0103] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in the present application.
[0104] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.
Claims
1. A model training method, characterized in that: The method comprises: Display the processing unit labels corresponding to each processing unit involved in the model training, display the input connection points and output connection points included in each processing unit label, and display the parameter item setting controls included in the customizable parameter processing unit label; the customizable parameter processing unit label is a processing unit label for which the user can customize the parameter items; Based on the operation of the user connecting the output connection point of at least one first processing unit tag to the input connection point of another first processing unit tag, and / or the parameter item information set by the user in the parameter item setting control, a first model training link is obtained; the first processing unit tag is the processing unit tag required by the user to train the first model; Model training is performed based on the first model training link to obtain a first model.
2. The method according to claim 1, characterized in that The parameter item setting control includes at least one of a network structure switching control for a model to be trained, a format parameter setting control, and a function on / off setting control.
3. The method according to claim 1, characterized in that: After obtaining the first model training link, the method further includes: Adjusting the first model training link to obtain a second model training link; Model training is performed based on the second model training link to obtain a second model.
4. The method according to claim 3, characterized in that The first model training link is adjusted to obtain a second model training link, including: In response to the user's tag adding operation, tag deleting operation, or tag order adjustment operation, adjusting the first model training link to obtain a second model training link; And / or, in response to the user's parameter item information adjustment operation, the first model training link is adjusted to obtain the second model training link.
5. The method according to claim 1, characterized in that Before displaying the processing unit labels corresponding to the processing units involved in the model training, displaying the input connection points and output connection points included in each processing unit label, and displaying the parameter item setting controls included in the customizable parameter processing unit label, the method further includes: Determine the data type of each input connection point and each output connection point; Get the target format standard corresponding to each data type; According to the target format standard corresponding to each data type and the data type of the input connection point and the output connection point of each processing unit tag, the input connection point and the output connection point of each processing unit tag are format-standardized.
6. The method according to claim 1, characterized in that Based on the operation of a user connecting an output connection point of at least one first processing unit label to an input connection point of another first processing unit label, a first model training link is obtained, including: Arrange the first processing unit labels based on the sequence of the processing processes required for training the first model by the user to obtain arranged first processing unit labels; Based on the user's operation of connecting the arranged target processing unit labels according to the input connection points and the output connection points of the first processing unit labels, a first model training link is obtained.
7. The method according to claim 1, characterized in that The performing model training based on the first model training link to obtain a first model includes: Based on the connection order of each first processing unit tag in the first model training link, calling the pre-stored encapsulated code segment corresponding to each first processing unit tag to obtain the model training code corresponding to the first model training link; Model training is performed based on the model training code corresponding to the first model training link to obtain a first model.
8. A model training device, characterized in that: The device comprises: A processing unit label display module is used to display the processing unit labels corresponding to each processing unit involved in the model training, and to display the input connection points and output connection points included in each processing unit label, and to display the parameter item setting controls included in the customizable parameter processing unit label; the customizable parameter processing unit label is a processing unit label for which the user can customize the parameter items; A first model training link acquisition module is used to obtain a first model training link based on an operation of a user connecting an output connection point of at least one first processing unit tag to an input connection point of another first processing unit tag, and / or parameter item information set by the user in the parameter item setting control; the first processing unit tag is a processing unit tag required by the user to train a first model; The first model training module is used to perform model training based on the first model training link to obtain a first model.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.